Google Project Genie lets you create interactive worlds from a photo or prompt

Google has unveiled Project Genie, a groundbreaking AI tool that allows users to craft immersive interactive environments from either a photograph or a simple text prompt. This development builds on last year’s Genie 3, which showcased remarkable long-term memory capabilities in generating dynamic worlds. Initially, Genie 3 was accessible only to a select group of testers. Now, Project Genie is open to a broader audience, though it remains exclusive to subscribers of Google's premium AI service. The concept behind world models, like Genie, is to create a living, responsive environment that adapts to user interactions, albeit not in a traditional 3D sense. Unlike standard virtual environments, Genie 3 introduces a unique feature where it can recall elements of the world it constructs for a limited duration. Although this “long time” is typically just a few minutes, it enhances the realism of the generated content. Project Genie is an enhanced version of Genie 3, integrating advanced AI models such as Nano Banana Pro and Gemini 3. Within Project Genie, users can explore various pre-built worlds, but the true allure lies in its customization capabilities. By submitting a reference image or providing a description of desired elements, users can initiate what Google refers to as “world sketching.” The AI first creates a still image based on user input, which can then be transformed into a fully interactive world. Should users find the initial reference image unsatisfactory, they have the flexibility to modify it before passing it to Genie for further development. The final output is a 720p video, rendered at approximately 24 frames per second, allowing for a nearly real-time experience as players navigate their characters using standard controls like WASD.